A Query For The Program Of Tribe-Mcl

Question

[Ontheway]

Hi, I am doing some work about clustering of proteins.But there is a problem:the clustering of multi-domain sequences,which most of the clustering methods can not solve. Tribe-MCL algorithm is one of the few methods which can deal with this problem.Unfortunately,I have not found the source code. If you have the source code or know the download website,please send me a copy.My email address is '[e-mail]'. Thank you !!!

[Michael]

Did you try to contact the authors?

Answer 1

[Micans]

Author of MCL here. The 'tribe' part of Tribe-MCL is the layer outside the clustering algorithm that does data preparation and mapping of labels, in the context of protein sequence similarity analysis. This functionality has for quite a while now been natively supported in the MCL software. You can get the MCL software from:

http://micans.org/mcl/

Instructions for dealing with BLAST results et cetera are provided in the 'protocols section', or, as a separate webpage:

http://micans.org/mcl/man/clmprotocols.html#blast

Hope this helps

[Ontheway]

Thank you. And I have another question: Can MCL solve the problem of multi-domain sequences clustering?

[Micans]

The answer is no I guess. MCL does fairly well at the multi-domain sequence problem, but it is completely oblivious to the domain structure; it knows only about the BLAST scores. This means that MCL is able to utilise those BLAST scores relatively well (and I am sure other cluster algorithms do too), but that is not 'problem solved'. In a way, BLAST scores are but a 'shadow' of sequence similarity space, and a method that knows more than just BLAST scores should be able to do better. Also, sequence 'families' are an approximation, I would say that orthology/paralogy are closer to reality.

[Ontheway]

Thanks for your help.^_^